Variant effect prediction (VEP)

We used PAINTOR software⁶⁹ to prepare the set of SNPs for VEP annotation. For this analysis, we provided PAINTOR with clumping results, LD matrices and annotation files calculated using the same 10,000 UKB individuals reference set that we used for COJO analysis. With PLINK⁷⁰ and we performed clumping analysis with ‘p1’ and ‘p2’ p-value threshold parameters set to 5 × 10⁻⁸, ‘r2’ set to 0.1 and MAF > 0.002. Then, we generated pair-wise correlation matrix for all SNPs in each region in clumping analysis results using plink–r option. When running PAINTOR, we did not use annotations; we changed options controlling input and output files format only, and otherwise we have used default parameters. We choose 159 SNPs marked by PAINTOR as 99% credible set for further analysis. In the next step, each SNP was extended with a list of proxy SNPs with R2 > 0.8 calculated using EUR cohort from 1000 Genomes Project Phase 3⁷¹ (N = 503) with 84.4 million variants as reference set. Total 924 SNPs was chosen for functional annotation by VEP with GRCH37 genomic reference.